February 26-28, 2025
Montreal, Canada
From Idea to Impact: building RAG beyond proof-of-concept
Retrieval-Augmented Generation (RAG) is a technique that helps reduce LLM hallucinations by grounding responses in external, accurate data sources. However, taking a RAG beyond PoC is not easy. In this talk you'll learn about the challenges of naive RAG implementation, and what can help you build realistic RAG. We’ll cover topics like re-ranking, metadata filtering, BM25 and hybrid search, and evaluating RAG systems beyond a “vibe check”.
Maria Khalusova
Unstructured.io
Maria Khalusova, a Staff Developer Advocate at Unstructured.io, currently works on addressing the challenges of preprocessing complex unstructured data for Generative AI applications, such as Retrieval Augmented Generation. Previously, she contributed to the field through her work on the open-source team at Hugging Face. Maria has co-authored a course on building applications with open source models, and a course on Audio Transformer models, both of which are openly available to all.
